Research map / Vision-language and multimodal / Visual question answering
Exploiting Background Knowledge in Compact Answer Generation for Why-Questions
Ryu Iida, Canasai Kruengkrai, Ryo Ishida, Kentaro Torisawa, Jong-Hoon Oh, Julien Kloetzer
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add word labels for causal-relation cause parts (ablation) | Proposed+AvgInit · Compact why-answer test set | ROUGE-1 F: 54.75 -> 55.15 | Improved |
| add stacked CNNs in encoder (ablation) | Proposed+AvgInit · Compact why-answer test set | ROUGE-1 F: 54.62 -> 55.15 | Improved |
| add decoder initialization with averaging (AvgInit) (ablation) | Proposed · Compact why-answer test set | ROUGE-1 F: 54.83 -> 55.15 | Improved |
| add causal-relation word labels plus stacked CNNs vs Ishida-LSTM | Ishida-LSTM · Compact why-answer test set | ROUGE-1 F: 52.33 -> 54.83 | Improved |
| add query-attention using causal relations | Ishida-GRU · Compact why-answer test set | ROUGE-1 F: 52.82 -> 52.82 | No clear effect |
| add key-value memory network using causal relations | Ishida-LSTM · Compact why-answer test set | ROUGE-1 F: 51.68 -> 51.96 (R-L) | No clear effect |
| add KVMemNN to Proposed+AvgInit | Proposed+AvgInit · Compact why-answer test set | ROUGE-1 F: 55.15 -> 54.71 | Hurt |
| use background knowledge instead of adding auto-generated training data (250K) | Ishida-LSTM · Compact why-answer test set | ROUGE-1 F: 54.51 -> 55.15 | Improved |
Closest papers by meaning
- Semi-Distantly Supervised Neural Model for Generating Compact Answers to Open-Domain Why Questions (AAAI 2018 · similarity 0.90)
- Improving Question Generation with Sentence-Level Semantic Matching and Answer Position Inferring (AAAI 2020 · similarity 0.86)
- SuRe: Summarizing Retrievals using Answer Candidates for Open-domain QA of LLMs (ICLR 2024 · similarity 0.85)
- RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs (NeurIPS 2024 · similarity 0.84)
- RAG-IGBench: Innovative Evaluation for RAG-based Interleaved Generation in Open-domain Question Answering (NeurIPS 2025 · similarity 0.84)
- Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
